It is desirable to detect from values x[t] the presence/absence of an information signal s[t] within ‘noise’ n[t] i.e. whether x[t]=n[t] or x[t]=s[t]+n[t].
This may be difficult if the signal to noise ratio s[t]/n[t] is small. It also becomes difficult to find a single algorithm that can deal with large ranges of signal to noise ratio.